Kerri Pickel. 21st Century Psychology: A Reference Handbook. Editor: Stephen F Davis & William Buskist. Volume 1. Thousand Oaks, CA: Sage Publications, 2008.
According to the Innocence Project (www.innocenceproject.org), created by Barry Scheck and Peter Neufeld, approximately 75 percent of the wrongly convicted people who have been eventually freed from prison through forensic DNA testing were found guilty at trial because of mistaken eyewitness identifications. The results of many studies have demonstrated that witnesses can make errors when they try to describe perpetrators or identify them in lineups. However, jurors tend to trust witnesses’ accounts and give them substantial weight when making decisions about defendants’ guilt. Studying the variables that affect the accuracy of eyewitness memory is important because the knowledge gained could lead to better techniques for questioning witnesses and more effective evaluations of witnesses’ statements.
In the 1970s, Elizabeth Loftus began publishing ground-breaking studies showing that questioning witnesses in certain ways can change their reports. Her work inspired numerous other researchers to examine various aspects of eyewitness performance. Most eyewitness memory studies are laboratory experiments, which have the advantage of allowing researchers to control extraneous variables and confounds and thus improve the chances of discovering causal relationships. The participants are often college students, although sometimes community members are recruited. The participants act as witnesses by viewing a video, a slide sequence, or a live, staged event enacted by confederates. The researchers may disguise the purpose of the experiment so that, while viewing the critical event, the witnesses are unaware that they will later have to try to remember it. In this way, researchers hope the participant-witnesses’ experience will be similar to that of actual witnesses, who are often simply going about their daily activities when they are surprised to notice a crime unfolding in their vicinity.
After watching the critical event, the witnesses may complete a written questionnaire that asks them to remember what happened and to describe the characteristics of the perpetrator. Alternatively, they may be interviewed by the experimenter, with the interview recorded for later analysis. In addition to or instead of asking for descriptive information about the event or the perpetrator, the witness could be required to view a lineup in which they try to identify the guilty person. Most lineups are photo lineups, in which witnesses examine a set of mugshot-type photos, but it is also possible to use a live version so that the lineup members are physically present. As in the real world, the true perpetrator may or may not appear in the lineup along with the innocent distracters (called fillers or foils). If he or she does, it is a target-present lineup. When a witness correctly identifies the target, it is called a hit. Target-absent lineups are created in the lab by replacing the perpetrator with a foil who resembles him or her. A false alarm occurs when a witness incorrectly identifies a foil as the perpetrator. As the Innocence Project has shown, it is this error that can send individuals to prison for crimes they did not commit.
In a laboratory experiment, the researcher can manipulate one or more independent variables to observe the effect on the witnesses’ memory. The independent variable could relate to some aspect of the critical event (e.g., whether the perpetrator is carrying a weapon), the wording of the questions asked after the event, or the way the lineup is constructed. In this way, researchers can try to determine whether these variables causally influence memory. Although experiments can be quite effective at accomplishing this goal, they could be criticized for having low external validity. In other words, concerns could be raised about whether the results would generalize to actual crime situations due to certain characteristics of the experiments. For example, college student participants may differ in important ways from typical witnesses to real crimes, experimental witnesses may feel less anxiety during the critical event, and they know that their reports will have no real consequences.
An alternative research design that obviously has higher external validity is the field study, which typically uses an archival approach to analyze the reports or lineup identifications made by real witnesses. Although the external validity is good, there are two related disadvantages of this type of design. First, there are usually multiple uncontrolled variables that can make it difficult to determine cause and effect. For example, across cases there may be variations in the duration of the crime event, the physical distance between the witnesses and the perpetrators, the elapsed time between the event and the witness’s report, the interviewing techniques used by police, and lineup instructions. Second, when studying actual crimes it is difficult (and sometimes impossible) to figure out exactly what happened, who committed the crime, and how the perpetrator looked at the time, and this information is clearly needed to assess the accuracy of witnesses’ reports. Even when someone is arrested and convicted, researchers cannot be sure that this person was the perpetrator, as the Innocence Project’s data demonstrate. Moreover, when it can be established that a certain person is guilty, it is still hard to know exactly what he or she did during the crime or whether the witness provided an accurate description of his or her appearance at the time.
Because lab and field studies have different strengths, they can be seen as complementary, and researchers can use both to learn about eyewitness memory. Together the two types have revealed a number of variables that affect witness accuracy. Researchers have found it convenient to classify these variables into two categories: system and estimator variables. The former are so named because they can be controlled by the criminal justice system. For instance, police investigators can choose the instructions that they give to witnesses preparing to view a lineup. Estimator variables can be estimated but not controlled by the system. An example of this type of variable would be whether the witness happens to be of the same race as the perpetrator. It is not within the scope of this article to describe in detail all of the many variables that researchers have linked to eyewitness performance, but the next sections will review what is known about some in each category.
Effects of Stress
It seems logical to assume that witnessing a crime would be stressful, particularly if the perpetrator has a weapon or commits violence. It is difficult to know for sure how witnesses feel during a real crime, however, because the best that researchers can do is to ask these witnesses to self-report their emotional state at some time after the event. Retrospective reports can be inaccurate because of witnesses’ imperfect memory of their prior state (especially after longer delays) and biases such as understating of the level of anxiety felt during the crime due to a motivation to minimize the dangerousness of the experience. Despite the measurement problems, researchers have asked real-life witnesses to rate their anxiety during the crime. The findings show differences among witnesses, some of which can be attributed to various aspects of the crime situation, such as the witnesses’ closeness to the perpetrator and the duration of the event. Not all witnesses report feeling terrified. Nevertheless, some clearly experience more stress than they would during a normal daily event, and therefore it is important for researchers to explore the effects of stress on memory accuracy.
A starting point is to define what is meant by stress. Sometimes the terms stress, anxiety, arousal, and negative emotional stress are used almost interchangeably within the literature, and these terms confuse physiological and emotional reactions. However, researchers have begun to distinguish between two types of responses to stimuli that are ostensibly stressful. One is a defensive physiological response that occurs when the activation mode of attentional control dominates. It involves increases in heart rate and blood pressure as well as vigilance and readiness for action (e.g., escape). This type of response can be triggered by stimuli that cause cognitive anxiety or worry as well as the perception of physiological arousal. The second type is an orienting response that occurs when the arousal mode of attentional control is more active. Here, heart rate and blood pressure are lowered, and attention is directed to stimuli that seem informative or novel to the individual. These stimuli could have negative emotional aspects to them, but not necessarily. For the present purposes, a stressful stimulus is one that produces a defensive response, and a negative emotional stimulus will be defined as one designed to cause witnesses to feel some sort of negative emotion (e.g., sadness, revulsion), probably without producing a defensive response, as discussed below.
At one time researchers generally argued that stress causes witnesses to allocate their attention to the “central” details of a scene (details connected with the theme or gist of the event) at the expense of “peripheral” details (irrelevant to the gist). For example, in one representative lab experiment, Christianson and Loftus (1991) asked witnesses to watch a slide sequence in which the critical slide showed either a neutral event (a woman riding a bicycle) or a negative emotional event (the aftermath of a car-bicycle accident in which the woman was injured and shown lying on the ground, bleeding from a head wound). The witnesses subsequently remembered central information from the critical slide (the color of the victim’s coat) more accurately if the slide was emotional rather than neutral, but the reverse was true for the peripheral information (the color of an uninvolved car in the background). Results like these were usually explained by referring to the Yerkes-Dodson law and J. A. Easterbrook’s (1959) cue-utilization hypothesis. According to the Yerkes-Dodson law, stress predicts performance as an upside-down U function, so that memory accuracy declines as stress increases above the optimal level in the center of the curve. Easterbrook’s hypothesis was used to explain in more detail the cognitive processes that occur: As stress increases, the range of stimuli that an individual can attend to becomes narrower, and consequently the individual must allocate his or her attention to the most informative or useful stimuli in the environment (i.e., the central details) instead of other stimuli (i.e., the peripheral details).
More recently, some researchers have reinterpreted the studies that seemed to demonstrate that stress improves memory for central details while harming memory for peripheral information. In their review, Deffenbacher, Bornstein, Penrod, and McGorty (1994) argued that these studies elicited an orienting response rather than a defensive response. In other words, they thought that, although the emotional stimuli attracted the witnesses’ attention (perhaps because of their salience, unusualness, or importance to the story), they did not actually cause stress. If it is true that witnesses allocated considerable attentional resources to the emotional stimuli, it would make sense that they later remembered less about other details than they otherwise would have.
Obviously, to determine the effects of stress on memory, one must look at studies in which stress was truly manipulated or varied naturally across conditions. Deffenbacher et al. (1994) contended that such studies do exist, but they have sometimes been confused with ones that merely include negative emotional stimuli. A set of studies by Douglas Peters (1988, 1991, 1997) serves as an example. In them, the stressful stimulus was a medical inoculation that participants received. In a meta-analysis that included Peters’s studies and others, Deffenbacher et al. reported that high levels of stress impaired both lineup identification and descriptions of target persons (if measured by answers to specific questions asked during an interview but not free recall). With respect to lineups, the authors speculated that high stress might degrade the witness’s memory representation of the perpetrator so that it becomes difficult to match that representation to him or her in a target-present array but easy to see similarities between the memory representation and one of the foils in a target-absent array. Deffenbacher et al. also suggested that the impaired memory representation of the target might be less apparent when free recall is used instead of interrogation because in the former case the witness can decide what to include, what to exclude, and in what order to report the information.
Exactly how are stress and memory accuracy related? To answer this question, Deffenbacher et al. (1994) appealed to a “catastrophe model.” In contrast to the Yerkes-Dodson curve, which predicts that as stress increases above the optimal level memory performance will gradually drop, the catastrophe model proposes that, at some critical point, the drop will be sudden and discontinuous if stress reactions have become too extreme. As a result, memory for the stressful event will be weak.
Biologically, stress causes the adrenal glands to release glucocorticoids. If their levels are high enough, these hormones can disrupt the normal functioning of the hippocampus, which is a brain structure instrumental in organizing and consolidating episodic details into coherent long-term memories. Although some aspects of the overall gist of a stressful event may be reasonably well remembered thanks to the involvement of another brain structure called the amygdala, the memory representation may lack spatial, temporal, and other contextual details if the hippocampus is not working normally. Given the discussion thus far, it may seem puzzling that individuals who have experienced an extremely stressful event (one we might call traumatic) often have memories of that event that are vivid and intrusive, so that unwanted flashbacks may occur. However, traumatic memories are not necessarily accurate, and sometimes details of one event are blended with those of another similar experience, or the account includes details that did not happen at all but instead were simply imagined or thought about. In addition, traumatic memories are often fragmented and lacking in contextual detail, as when a soldier recalls mainly a mental image of a dead comrade’s face when remembering a combat episode.
In sum, empirical evidence suggests that negative emotional stimuli most likely attract attention by causing an orienting response, but witnesses probably do not experience stress when viewing them. These stimuli will receive favored processing relative to neutral stimuli, but the advantage is restricted to central details related to the emotional stimuli as opposed to peripheral or background information. Stressful stimuli, on the other hand, seem to produce a defensive response that results in a general impairment of memory except perhaps the overall gist.
The Weapon Focus Effect
Witnesses who observe an armed perpetrator tend to direct their attention toward the weapon, and subsequently they are unable to remember information about the perpetrator’s physical appearance as accurately as they would have if no weapon had been visible. This phenomenon is known as the weapon focus effect. Many experiments have demonstrated that a weapon’s presence reduces the accuracy of witnesses’ descriptions of the perpetrator and their ability to identify that person in a lineup, although the effect size with the latter variable is weaker, probably because it is a less sensitive measure of memory.
A good illustration of the effect is found in a study conducted by E. F. Loftus, G. R. Loftus, and Messo (1987). Their witnesses viewed a slide sequence depicting customers moving through the order line at a fast-food restaurant. The perpetrator, who was shown on four critical slides, held either a gun or a personal check as he took money from the cashier. Witnesses who saw the gun were less likely than those who saw the check to identify the perpetrator in a photo lineup (in two experiments), and their descriptions of him were less accurate (in one experiment). E. F. Loftus et al. used a corneal reflection device to track witnesses’ eye movements as they watched the slides. They found that the witnesses fixated on the gun more frequently and for longer durations than they fixated on the check. This result, along with the finding that the weapon did not affect witnesses’ statements about details presented before the critical slides, implies that weapons both attract attention and decrease witnesses’ ability to encode other details that appear simultaneously with the weapon.
In addition to E. F. Loftus et al.’s (1987) study, many other lab experiments have obtained the weapon focus effect. Various types of stimuli have been used (slides, video, and live, staged events), as well as different weapons (e.g., handgun, knife, meat cleaver, liquor bottle, and syringe). Furthermore, community samples as well as college students have participated.
Among field investigations of the effect, the findings are mixed, with many (but not all) reporting null results. Remember, field studies are complicated by the presence of uncontrolled variables that could influence witnesses’ reports and obscure the causal relationship between the weapon and memory for the perpetrator, if this connection exists. For example, an armed criminal might feel emboldened to approach nearer to the witness or stay at the scene longer than an unarmed one, which in turn could cause the witness to encode information about the criminal more completely and more deeply, thus mitigating the weapon focus effect. Similarly, the police might regard crimes involving a weapon to be more serious, so they might question witnesses to these crimes more thoroughly and therefore elicit better descriptions. More research is needed to clarify the field results.
Among the possible explanations for the weapon focus effect, the most intuitively appealing may be that a weapon causes witnesses to feel frightened or anxious due to the possibility that they or someone else could be injured or killed. As their anxiety rises, their attentional capacity shrinks so that they become unable to encode as many details simultaneously as they normally would. Therefore, they focus on details that are considered important, such as the weapon, which is the source of the anxiety. Less important details, such as the color of the perpetrator’s shirt, receive minimal processing. This explanation, which is based on Easterbrook’s (1959) cue-utilization hypothesis, is opposed by several results. First, the weapon focus effect has been obtained even when witnesses’ rated anxiety levels are low. This result suggests that an orienting response rather than a defensive response may be at work, and that weapons do not necessarily cause genuine stress. Additionally, witnesses’ memory for the perpetrator is not harmed by the degree to which the perpetrator threatens a victim or the degree of threat associated with the object held by the perpetrator.
A better explanation is based on unusualness: Weapons are unexpected within many contexts. For example, even though shoppers know that convenience stores sometimes get robbed, the probability of witnessing a robbery (and seeing a gun) during any given visit to one of these stores is low. Moreover, previous attention research has shown that people look more often and for longer durations at unusual rather than typical objects within visual scenes. In support of this hypothesis, researchers have found that an unusual object (e.g., a stalk of celery or a raw chicken) can reduce memory accuracy just as a weapon can. Also, in accordance with the hypothesis, the weapon focus effect disappears if the witness sees the weapon in a context in which it would be expected such as a shooting range.
Recently researchers have tried to determine whether witnesses’ fixations on a weapon are automatic (and thus involuntary and unavoidable) or controllable. This issue is interesting for both theoretical reasons, because it has implications for researchers’ ability to understand witnesses’ attentional processes, and practical reasons, because it speaks to the possibility of training potential witnesses to provide more accurate reports. Pickel, Ross, and Truelove’s (2006) participants listened to a lecture that either informed them about the weapon focus effect, instructing them to try to avoid focusing on a weapon, or covered an unrelated topic. Subsequently the participants witnessed a staged event involving a perpetrator who carried either a gun or a textbook. The witnesses who heard the control lecture exhibited the usual weapon focus effect, describing the perpetrator more accurately when he held the book instead of the gun. In contrast, the educated witnesses remembered the perpetrator equally well in the two object conditions. From these results, Pickel et al. concluded that weapons probably do not capture attention automatically, and witnesses can overcome the effect if informed about it.
The Own-Race Bias
Witnesses can more accurately remember faces of members of their own racial group than those of another group. Meissner and Brigham (2001) conducted a meta-analysis that reviewed 91 different studies in an effort to find overall statistical conclusions about this effect. They discovered that the own-race bias is apparent when both hits and false alarms are used as dependent measures. Across the studies they reviewed, witnesses were 1.4 times more likely to make a correct lineup identification if the target was of the same rather than a different race, and they were 1.56 times more likely to make a mistaken identification. They also discovered that, although witnesses of any race can display the own-race bias, the effect was larger for white than for black witnesses when considering false alarms (but not hits). Furthermore, Meissner and Brigham reported that false alarms for other-race faces increased when witnesses viewed the target for a shorter time. Additionally, follow-up research has demonstrated that witnesses are highly confident in their mistaken identifications of other-race targets and that children as young as kindergarten age show the same bias that adult witnesses do.
Several social and cognitive processes have been invoked to interpret the own-race bias. One proposal is that witnesses with prejudiced racial attitudes are not very motivated to differentiate between other-race individuals. However, most recent studies have not obtained a correlation between attitudes and other-race memory performance. Researchers have also examined the possibility that the faces of members of some racial groups might show less physical variability than others, which would make those faces hard to distinguish. As with the first proposal, little support was found for this hypothesis. The most promising explanation of the other-race bias draws upon results showing that witnesses who have had more contact with members of other races are less susceptible to the effect. For example, in one study, children and adolescents living in integrated neighborhoods could better recognize other-race faces than those living in segregated areas could. Moreover, several data sets have indicated a positive correlation between witnesses’ self-reported level of contact with other-race individuals and the accuracy of their lineup identifications of other-race targets (Meissner & Brigham, 2001).
Assuming that interracial contact matters, the next step for researchers is to determine why. Although decreases in prejudice and stereotypic beliefs could play a role, most accounts focus on perceptual learning, or the idea that increased contact helps one to learn more about the physical features associated with other-race faces and to memorize them better. Specifically, researchers have suggested that, with experience, individuals learn to identify and attend to the features that are most relevant to the task of discriminating one face from another. Over time, other-race faces can be encoded in memory more efficiently and with a richer, more complete representation.
Even with the possibility that individuals may be able to reduce their susceptibility to the own-race bias through interracial contact, the effect remains problematic for the legal system. As Meissner and Brigham (2001) noted, it is especially troubling that mistaken identifications are more likely if the witness and target are of different races because this kind of error can lead to a prison term for an innocent person.
In the 1970s E. F. Loftus began publishing research demonstrating that eyewitness memory reports can be affected by postevent information, or information that the witness encounters after the target event. E. F. Loftus showed that, if the postevent information is misleading, witnesses may subsequently report details that are consistent with that information instead of correct details. This phenomenon is known as the misinformation effect.
For example, consider an early experiment conducted by E. F. Loftus, Miller, and Burns (1978) in which witnesses viewed a slide sequence depicting a traffic accident. In one condition, the critical slide showed a car stopped in front of a yield sign at an intersection. Afterward the witnesses completed a questionnaire about the slides. The actual purpose of the questionnaire was not to test witnesses’ memory but to manipulate the independent variable, which was the presence of misleading information. One version of the questionnaire contained the critical item “Did another car pass the red Datsun while it was stopped at the stop sign?” This item is misleading because it incorrectly identifies the type of sign. The control version of the questionnaire included the same item with the words “yield sign” replacing “stop sign.”
After a 20-minute filler activity, the witnesses tried to identify the critical slide they had seen in a two-alternative recognition test. As would be expected, most (75 percent) of the control witnesses chose the correct slide. However, most (59 percent) of the misled witnesses chose the slide that was consistent with the misinformation they read instead of the slide they actually viewed.
In this experiment witnesses erred by substituting a suggested detail for a correct detail. E. F. Loftus also showed in other studies that witnesses can also be induced to report seeing nonexistent objects. For example, witnesses who had watched a video of an automobile accident answered a series of questions about it. Some witnesses were asked, “How fast was the white sports car going when it passed the barn while traveling along the country road?” This item is misleading because no barn appeared in the video. The control version of the question omitted the phrase “when it passed the barn.” After a delay of one week, all witnesses completed a memory test that included the question “Did you see a barn?” Although most witnesses correctly answered “no,” 17 percent of the misled witnesses reported seeing a barn, compared to only 3 percent of the controls.
The misinformation effect has been replicated many times by many researchers. The misleading information may be embedded in questions, narrative accounts, or pictures presented to the witnesses. Various kinds of target events have been used, as well as delay intervals of differing lengths. Across these studies, the typical result is that a substantial proportion of misled witnesses provide reports that are consistent with the misinformation they encountered. Sometimes witnesses will even provide descriptive details about nonexistent objects whose presence was suggested to them.
Researchers have proposed several hypotheses about the cause of the misinformation effect. In her early work, E. F. Loftus suggested that the misleading information might alter or replace the witnesses’ memory trace of the target event so that the original trace no longer existed. This “overwriting” hypothesis is probably too simplistic, however, because it turns out that in some situations misled witnesses can retrieve the original, correct information. Partly in response to this finding, some researchers have argued that the misinformation effect is not the result of memory impairment but instead reflects strategic choices made by misled witnesses. For example, it is possible that witnesses can remember the target event perfectly well, but they realize that the experimenter expects them to report details consistent with the misinformation. Therefore, in an effort to be cooperative, they report what the experimenter wants to hear instead of what they actually remember. Besides this type of demand effect, another way that a strategic effect could emerge is if misled witnesses cannot recall the original details but assume that the misinformation must be accurate because it was given to them by the experimenter. This result is termed “misinformation acceptance.” Witnesses’ strategic choices probably account for some instances of the misinformation effect, but they cannot explain some data. For example, in several studies misled witnesses were asked during the memory test phase of the experiment not to report any details from the target event. Often these witnesses avoided reporting the misinformation, which implies that they truly believed that these details were part of the target event.
Two other explanations have some empirical support. First, the retrieval-blocking hypothesis says that, at the time of the test, memory traces for both the target event and the misinformation exist and are in competition with each other. If the misinformation trace is stronger (for example, because those details were encountered more recently), then that trace will not only be more accessible but will also block access to the accurate trace.
A final explanation is based on source monitoring. Many studies have demonstrated that people sometimes forget the source of information they have encountered (or they failed to encode the source in the first place). Consequently, they may misattribute the source. In other words, witnesses might remember the misinformation but wrongly believe those details were part of the target event rather than the postevent questionnaire. Consistent with this explanation, some data indicate that asking witnesses to pay special attention to the sources of the information they receive reduces suggestibility.
False Memories and the Effects of Imagining
Building on research on the effects of misleading information, researchers have found that it is possible to implant in witnesses false memories for events that never happened. For example, E. F. Loftus and Pickrell (1995) asked college students to read descriptions of events written by a relative (e.g., a parent or an older sibling). Some of the described episodes really happened to the participants, but one in each set was a false event about getting lost in a shopping mall as a small child, and how the participant cried before being found by an elderly woman and reunited with relieved family members. The participants initially wrote down any memories they had of the events. Then, they were interviewed twice during a four-week period, each time trying to recall as much as possible about the events while using their relatives’ descriptions as cues. This procedure induced 25 percent of the participants to say during the final interview that they remembered the false event (by comparison, 68 percent of the true events were remembered), although they rated the false event as less clear than the true ones, and they were less confident about those memories.
It appears that asking participants to imagine an event is a fairly effective method of creating a false memory. Garry, Manning, E. F. Loftus, and Sherman (1996) found that instructing participants to visualize a childhood event and guiding them as they imagined different aspects of it significantly increased their ratings of the likelihood that the event occurred. Other studies suggest that people who are skilled at forming clear mental images may be especially susceptible to creating false memories, and imagining can increase people’s belief that they can remember experiencing an event, including contextual details, instead of simply becoming more convinced that they know the event occurred. Researchers have proposed that imagining leads to the creation of plausible and vivid details associated with the event, and later participants may become confused about the source of those details, believing they were part of the actual event.
A potential criticism of studies of false memories is that perhaps the researchers have assumed that the “false” event did not occur in a participant’s life, but actually it did. Studies designed to address this issue have asked participants about events that could not possibly have happened, such as medical procedures that never occur in the participants’ home country, with the same results as previous studies.
Because of the abundance of data showing how false memories can be created, researchers have cautioned against the use of imagining as a tool to help retrieve memories. For example, police investigators have sometimes asked witnesses to imagine in an effort to assist them in remembering the details of a crime, and therapists have used guided imagery to help clients remember and deal with traumatic childhood memories. However, the data indicate that such techniques are likely to be counterproductive and could even lead people to “remember” episodes they never experienced.
After police investigators have identified a suspect, they may ask witnesses to view a photo lineup (or, less commonly, a live lineup) in an effort to determine whether the suspect is the perpetrator. Researchers have examined the relationship between a number of variables and identification accuracy. Some of those variables are reviewed below.
The Selection of Foils
A lineup includes one person whom the police suspect of committing the crime being investigated. This person may or may not be guilty. The other members of the lineup are not the focus of the investigation, and they are called fillers or foils, as noted previously. Researchers have discovered that the physical appearance of the foils can influence a witness’s choice. For example, if the suspect is innocent but resembles the perpetrator significantly more than the foils do, then the witness may mistakenly identify that suspect as the guilty party (a false alarm). Although selecting foils who look like the suspect may at first seem to be the fix for this problem, this strategy may lower the hit rate for target-present lineups because the lineup members may be so similar that they are indistinguishable. Studies show that it is actually better to use foils who match the witness’s description of the perpetrator.
Witnesses sometimes assume that a suspect arrested by the police is probably guilty and therefore that a lineup probably includes the perpetrator. They may also feel intense pressure to choose someone from a lineup in order to help solve the crime. Researchers studying police transcripts have found instances in which lineup administrators encouraged witnesses who had initially failed to identify anyone to take another look or pressured witnesses by telling them that they had to help get a dangerous criminal off the streets. However, even without explicit pressure from the administrator, witnesses may feel compelled to choose. To counteract this tendency, administrators can simply inform witnesses that the perpetrator may or may not be in the lineup. Studies show that this instruction significantly reduces false alarms in target-absent lineups without harming hit rates in target-present lineups.
After a witness makes an identification (or, alternatively, says that the perpetrator is not in the lineup), the police investigator, in an effort to determine whether the witness was accurate, may ask him or her for a rating or estimate of confidence. It seems reasonable to assume that highly confident witnesses are probably accurate and uncertain witnesses are probably wrong. Not surprisingly, jurors are profoundly influenced by a witness’s expression of confidence as they assess credibility during a trial. However, the strength of the actual correlation is affected by several variables such as whether the witness had a good opportunity to view the perpetrator and the degree of resemblance between the perpetrator and the innocent suspect (in the case of false alarms). Moreover, these variables can be difficult to estimate. As a result, from a practical viewpoint, the usefulness of confidence as a predictor of accuracy is limited.
A further problem with confidence is that it is malleable, meaning that witnesses’ confidence can be changed by exposure to information that may be unrelated to their performance in the lineup task. For example, Luus and Wells (1994) had pairs of witnesses watch a staged theft. After each witness independently completed a lineup task, the researchers told him or her that the other witness had either identified the same person, identified another person, or decided that the target was not present (there was also a control condition in which no information about the co-witness’s choice was provided). Finally, the witnesses rated their confidence. Telling witnesses that the co-witness had identified the same person caused an increase in confidence relative to the control condition, whereas telling them that the co-witness identified someone else or no one caused a drop in confidence. Interestingly, almost all of the witnesses in this study made a mistaken identification due to the procedures the researchers used. Luus and Wells argued that their findings could easily apply to real-life investigations because lineup administrators might tell one witness what another said, or the witnesses might exchange information directly if they come into contact with one another.
Another way to manipulate confidence is to give witnesses feedback on their own performance. It is not illegal for police investigators to give witnesses confirming feedback (e.g., telling them that they did a good job and identified the actual suspect) or disconfirming feedback (e.g., telling them that they chose someone who was not the focus of the investigation). However, this information might be incorrect because the police could have targeted an innocent person. Wells and Bradfield (1998) asked participants to watch a surveillance video that showed a man who later shot a security guard in a retail store. Afterward, the witnesses viewed a lineup and then were given randomly selected feedback. The authors found that confirming feedback inflated confidence relative to disconfirming feedback. Witnesses who received confirming feedback also said that they had a clearer view of the perpetrator, that they could more clearly see the details of his face, that they paid closer attention to the event, and that it was easier for them to make an identification. These witnesses, like those in Luus and Wells’s study, identified someone other than the actual perpetrator. However, jurors would no doubt find them quite convincing.
To reduce the likelihood that the lineup administrator will give feedback to the witness (either verbally or through nonverbal signals such as body language), researchers advise that a double-blind procedure should be used. In other words, the administrator should not be the police detective who is conducting the investigation of the case, but instead should be someone who does not know which lineup member is the suspect and which members are foils. Alternatively, if the lineup involves photos rather than live members, it could be presented by a computer. If one of these methods is used, there is less chance that the administrator could intentionally or inadvertently influence the witness’s confidence.
Simultaneous Versus Sequential Lineups
In the United States, most lineups include six members, or sometimes five. Traditionally, police investigators have presented all the members at once to witnesses, asking them to look at the entire lineup and decide whether the perpetrator is in it. However, researchers recommend using a sequential lineup instead. In the sequential procedure, witnesses are told that they will view an unspecified number of individuals one at a time. When a particular lineup member is shown, the witness must decide whether that person is the perpetrator. If the decision is “yes,” the procedure is finished. If the decision is “no,” the administrator moves on to the next lineup member. The process continues until the witness either identifies someone as the perpetrator or views all the members without identifying anyone.
The problem with the simultaneous lineup is that it encourages relative judgments. That is, witnesses compare the members and choose the one who most resembles the perpetrator. This strategy can lead to errors because there will always be one member who looks at least a little bit more like the perpetrator than anyone else, even if the resemblance is not strong. Therefore, if the lineup does not contain the guilty person, a false alarm is likely.
It is difficult for witnesses to make relative judgments when sequential lineups are used because the six members cannot be viewed at the same time and compared. As a result, witnesses tend to make absolute judgments, separately comparing each member to their memory of the perpetrator, and the number of false alarms decreases. A potential problem with sequential lineups is that they appear to reduce the number of hits as well as false alarms; in other words, these lineups induce witnesses to make fewer identifications, whether accurate or not. However, some researchers have argued that the sequential lineup merely weeds out lucky guesses without affecting the hit rate for witnesses who truly remember the perpetrator and recognize him or her in the lineup. In any case, researchers typically advocate the use of sequential lineups, largely to cut down on the chance of convicting innocent individuals.
It is clear that eyewitnesses can make mistakes, even if they are trying to be honest and cooperative. Their reports or lineup identifications can be distorted by stress, the presence of a weapon, exposure to misleading information, or suggestive questioning methods that encourage them to imagine what happened. They are less able to identify a perpetrator accurately if that person is of a different race. Several additional variables can affect their lineup performance or the confidence they express about their choice.
The results of the many eyewitness memory studies that have been conducted since the 1970s have allowed researchers to formulate a number of recommendations for police investigators and others who work within the legal system. Use of these recommendations could lead to better techniques for questioning witnesses and administering lineups, as well as a better understanding of the variables that limit eyewitness memory performance. Drawing upon empirical findings, researchers have developed procedures such as the cognitive interview (CI) to help police interviewers get as many accurate details as possible from witnesses while minimizing the amount of incorrect information reported. The CI includes several steps that the traditional police interview lacks. For example, the CI instructs interviewers to build rapport with witnesses as a way of putting them at ease and to let them control the pace and direction of the interview while refraining from interrupting them. Interviewers should also ask witnesses to “reinstate the context” that was present during the target event by recalling the physical environment (e.g., objects, smells, or sounds) and their emotions and thoughts at the time. These contextual details can serve as cues that help witnesses retrieve event details from memory.
Before the 1990s law enforcement personnel, prosecutors, and judges showed little interest in the eyewitness literature and in researchers’ suggestions. However, the situation changed when DNA testing began to exonerate innocent persons who had been convicted of crimes and the press began to publicize these cases. Realizing that there was room for improvement within the system, courts and police departments have become more willing to establish reforms based on scientific findings. In fact, in the late 1990s, Attorney General Janet Reno ordered the National Institute of Justice to oversee a committee of researchers, attorneys, and police officers who worked together to develop a set of national guidelines for interacting with eyewitnesses. A few U.S. jurisdictions have adopted these guidelines. In the future, perhaps even more will join in the effort to improve the performance of eyewitnesses and make sure that the evidence collected from them is used appropriately.